Vehicle track identification in synthetic aperture radar images

1. A computing device, comprising: a processor, a display; and memory that comprises instructions that, when executed by the processor, cause the processor to perform acts comprising: responsive to receiving a first coherent change detection image of a scene, calculating a Radon transform of at least a portion of the first coherent change detection image; identifying a peak of the Radon transform; identifying a vehicle track in the scene based upon the identified peak of the Radon transform; generating a second coherent change detection image by calculating an inverse Radon transform of the Radon transform; assigning a color value to a pixel of the second coherent change detection image, the color value corresponding to an attribute of the vehicle track; and displaying the second coherent change detection image on the display, wherein the second coherent change detection image captures the vehicle track.

2. The computing device of claim 1 , wherein identifying the vehicle track in the scene comprises identifying a pair of vehicle tracks, and wherein the pixel of the second change detection image corresponds to the pair of vehicle tracks.

3. The computing device of claim 1 , wherein identifying the vehicle track comprises identifying a pair of vehicle tracks, the acts further comprising: responsive to identifying the pair of vehicle tracks in the scene, computing a width of the pair of vehicle tracks based upon a separation of the peaks of the Radon transform; and outputting graphical data to the display indicating the width of the pair of vehicle tracks.

4. The computing device of claim 3 , wherein outputting graphical data indicating the width of the pair of vehicle tracks comprises: assigning a second color value to a second pixel of the second coherent change detection image corresponding to the pair of vehicle tracks; and displaying the second coherent change detection image with the second color value assigned to the second pixel.

5. The computing device of claim 1 , wherein identifying the vehicle track comprises identifying a first pair and a second pair of vehicle tracks, the acts further comprising: responsive to identifying the first pair of vehicle tracks in the scene, computing a width of the first pair of vehicle tracks based upon separation of the peaks of the Radon transform; assigning a second color value to a second pixel of the second coherent change detection image corresponding to the first pair of vehicle tracks; responsive to identifying the second pair of vehicle tracks in the scene, computing a width of the second pair of vehicle tracks based upon separation of the peaks of the Radon transform; comparing the computed width of the first pair of vehicle tracks and the computed width of the second pair of vehicle tracks; and responsive to the computed width of the second pair of vehicle tracks being different from the computed width of the first pair of vehicle tracks, assigning a third color value to a third pixel of the second coherent change detection image corresponding to the second pair of vehicle tracks.

6. The computing device of claim 1 , wherein generating the second coherent change detection image comprises performing morphological erosion on the inverse Radon transform.

7. The computing device of claim 1 , the acts further comprising: prior to identifying the peak of the Radon transform, applying a threshold mask to the Radon transform to generate a filtered Radon transform, wherein the threshold mask is based upon at least a minimum length of a vehicle track and a grayscale value of pixels corresponding to vehicle tracks in the first coherent change detection image, wherein identifying the peak of the Radon transform comprises identifying a peak of the filtered Radon transform, and wherein identifying the vehicle track in the scene based on the peak of the Radon transform comprises identifying the vehicle track in the scene based on the peak of the filtered Radon transform.

8. The computing device of claim 1 , the acts further comprising applying a mask to the first coherent change detection image before calculating the Radon transform, the mask based upon a height map of the scene.

9. A method comprising: receiving a first coherent change detection image of a scene; computing a Radon transform of at least a portion of the first coherent change detection image; identifying a peak of the Radon transform; identifying a vehicle track in the scene based upon the identified peak of the Radon transform; generating a second coherent change detection image of the scene by calculating an inverse Radon transform of the Radon transform; and assigning a color value to a pixel of the second coherent change detection image corresponding to the vehicle track; and displaying the second coherent change detection image with the color value assigned to the pixel.

10. The method of claim 9 , further comprising: segmenting the first coherent change detection image into a first plurality of pieces of a first equal size, wherein computing the Radon transform of at least a portion of the first coherent change detection image comprises: computing a Radon transform of each of the pieces of the first plurality of pieces, wherein identifying the peak of the Radon transform comprises: identifying peaks of each of the Radon transforms of the first plurality of pieces; computing an inverse Radon transform of each of the Radon transforms of the first plurality of pieces; forming the second coherent change detection image of the scene by assembling the inverse Radon transforms of the first plurality of pieces, wherein identifying the vehicle track in the scene comprises: identifying a vehicle track in the second coherent change detection image based upon the identified peak of each of the Radon transforms of the first plurality.

11. The method of claim 9 , further comprising generating the first coherent change detection image, wherein generating the first coherent change detection image comprises: receiving a third and a fourth coherent change detection image; and jointly preprocessing the third and fourth coherent change detection images to form the first coherent change detection image.

12. The method of claim 11 , wherein jointly preprocessing the third and fourth coherent change detection images comprises one of the following: performing principal component analysis on the third and fourth coherent change detection images; performing independent component analysis on the third and fourth coherent change detection images; computing the normalized coherence product of the third and fourth coherent change detection images; or computing the difference change product of the third and fourth coherent change detection images.

13. The method of claim 9 , further comprising applying a threshold mask to the Radon transform before identifying the peak, the threshold mask based upon a minimum pixel length of vehicle tracks in the first coherent detection image and a pixel intensity value of a pixel corresponding to a vehicle track in the first coherent detection image.

14. The method of claim 9 , wherein identifying the vehicle track comprises identifying a pair of vehicle tracks, the method further comprising: computing a width of the pair of vehicle tracks based on a separation between peaks of the Radon transform; and displaying an indication of the width of the pair of vehicle tracks.

15. The method of claim 9 , wherein identifying the vehicle track comprises identifying a first pair of vehicle tracks and a second pair of vehicle tracks, the method further comprising: computing a first width of the first pair of vehicle tracks based on a separation between peaks of the Radon transform; computing a second width of the second pair of vehicle tracks based on a separation between peaks of the Radon transform; and comparing the first width and the second width, wherein displaying the indication that the second coherent change detection image includes the vehicle track comprises depicting the first pair of vehicle tracks and the second pair of vehicle tracks in different colors in the second coherent change detection image based on their track width.

16. A computer-readable storage medium comprising instructions that, when executed by a processor, cause the processor to perform the following acts: responsive to receiving a first coherent change detection image of a scene, calculating a Radon transform of at least a portion of the first image; identifying a peak of the Radon transform; identifying a vehicle track in the scene based on an angle and distance parameters of the peak of the Radon transform; computing an inverse Radon transform of the Radon transform to create a second coherent change detection image; assigning a color value to one or more pixels corresponding to the one or more vehicle tracks in the second coherent change detection image; and displaying the second coherent change detection image.

17. The computer-readable storage medium of claim 16 , the acts further comprising: identifying a width of the vehicle track based upon the peak of the Radon transform; and displaying an indication of the width of the vehicle track in the second coherent change detection image.

18. The computer-readable storage medium of claim 16 , the acts further comprising generating the first coherent change detection image, wherein generating the first coherent change detection image of the scene comprises: receiving third and fourth coherent change detection images; and jointly preprocessing the third and fourth coherent change detection images to form the first coherent change detection image.